Fixing networking issues

[Cutriss]

So I've decided I want to take another swing at using my Echo. The 360 is nice, but it's clear to me there's some kind of networking (or otherwise) issue in play that makes the 360 occasionally have issues, and I'd like some help sorting it.

So right now, I have the following setup:

Ubiquiti EdgeRouter Lite (FiOS uplink and two network segments, does DHCP for both)

Segment 1 - just a simple Trend 8-port gigabit switch. Three connections - one to my DVR system (with InfiniTV PCI 4-tuner), one to the EdgeRouter, and one through a patch panel into the front room where my TV is. The total distance of that run is maybe 30 feet. The other end is a 360 Slim.

Segment 2 - the rest of my network, and can be omitted.

On my 360, tuning/watching recorded programs works fine, with one exception - after viewing a program for a while (usually 20-30+ minutes), I get the Network Error popup and the framerate drops.

The easy way I can fix this is to dismiss the error and then just either hit the Instant Replay button or otherwise navigate the stream (rewind, fast forward, etc). I find it surprising that whenever I try using FF here, I seem to be able to recover a few seconds to get back to "live". It's almost like there's a timing issue where I'm getting out of sync with the source and navigating the progress bar fixes it.

If I try to watch TV on the DVR PC directly, I have no issue at all - it stays synchronized.

The HTPC has an onboard Realtek 8111C gigabit NIC (the motherboard is an ASRock A780FullDisplayPort). I think I've read people in the past saying that they had issues with Realtek NICs. If so, what's the "golden standard" here that I should pursue?

Any other ideas?

[JohnW248]

Don't quite understand the purpose of the patch panel, but the extender and the MC Host should be connected to the trendnet switch.

But on the MC host, go to network and sharing and the change adapter settings and click on your network card and then properties. See what's bound to the Ethernet card, all the protocol etc. You might have something there which could be an issue.

If your FiOS router is handling DHCP for your network then there is just one gateway so what is the other segment? How is that set-up?

[Cutriss]

Don't quite understand the purpose of the patch panel, but the extender and the MC Host should be connected to the trendnet switch.

My house is wired - that's where the jacks in the living room terminate. I then have patch cables to go from the patch panel to the switch. Not sure what's confusing about that.

The only things connected to the TrendNet switch are the MC host, the extender, and the EdgeRouter (for Internet).

But on the MC host, go to network and sharing and the change adapter settings and click on your network card and then properties. See what's bound to the Ethernet card, all the protocol etc. You might have something there which could be an issue.

It's all bog-standard stuff. IPv6 is disabled. The default QoS driver is not (nor are the link-layer drivers). I haven't seen any guidance from anyone as to turning those off.

If your FiOS router is handling DHCP for your network then there is just one gateway so what is the other segment? How is that set-up?

I'm not using the FiOS router, the EdgeRouter is doing that, I'm wired directly into the ONT so I don't use the FiOS ActionTec router at all.

The EdgeRouter runs two different DHCP pools for each segment and handles routing between them.

[JohnW248]

So what happens if you pull the jumper from the router to the switch with the ETH and MC computer...does it work then? That would eliminate a lot of unknown variables and if you still have trouble it would tell us where to look.

[Venom51]

Diagnosing intermittent problems like this is very hard in an unmanaged network. You are simply plugging everything in and hoping it all plays nice together. If I was trying to diagnose what's happening and who the offender could be it would require the following.

2 x 100 meg hubs. Not switches actual old school hubs.
2 x additional PC's, preferably notebooks.

I would then place a hub in between the Media center PC and the switch. Plug the monitoring notebook in there as well. I would place the second hub in between the 360 and the switch along with a monitoring notebook there as well. Wireshark would be used on both monitoring notebooks to capture the entire conversation between the Media Center PC and the 360. Wait for the issue to happen and then stop both logging sessions noting the time the event occurred. Looking back through both sides of the conversation should clearly indicate who stopped talking first and provide some clues as to why. A managed switch with some SNMP traps and a receiver setup would also help you figure out what's happening. Could be something as simple as the NIC driver crashes and cause the port to renegotiate it's link with the switch. That could be a simple reason for what you are seeing.

I would also be noting the time of the event and looking through the event logs on the Media Center PC for clues.

In terms of good NIC interfaces you can never go wrong with an Intel NIC.

[josim]

I am not an expert on networking.
I have read here or AVS the the Trendnet switches have a small amount of buffer memory, 256KB, and that can cause the network error message you have been receiving. They advise one w/1Mb buffer.